Landmark based facial emotion recognition

Made by Wahbi Oussama 😤 and Oumayma 🙃

This project is an attempt at training a fully connected Neural Network to predict emotions using facial landmarks only. There are 7 emotions to predict

• Angry 😠
• Disgust 🤢
• Fear 😨
• Happy 😃
• Neutral 😐
• Sad 😢
• Surprise 😲

Project Setup

use this command to clone the repo in ur local setup:

git clone https://github.com/NullPointer-hash/landmark-based-facial-emotion-recognition.git

once you download the project, open CMD in the project directory or open CMD and use:

cd PROJECT_DIRECTORY

run this command to install all the required packages:

pip install -r requirements.txt

Gathering data for each emotion

python gather-data.py --help to show description of the parameters

use these 7 commands to gather data for each of the 7 emotions specify the number of iterations for each cmd

! it is a slow process if iterations are high !
!! emotion labels are case sensitive !!

python gather-data.py --iterations 100 --emotion Angry
python gather-data.py --iterations 100 --emotion Disgust
python gather-data.py --iterations 100 --emotion Fear
python gather-data.py --iterations 100 --emotion Happy
python gather-data.py --iterations 100 --emotion Neutral
python gather-data.py --iterations 100 --emotion Sad
python gather-data.py --iterations 100 --emotion Surprise

Training your own model

python train.py --help to show description of the parameters

you can train your own model using this command it is advice to use more than 6000 rows in the dataset to have good results

python train.py -e NUMBER_OF_EPOCHS
                -bs BATCH_SIZE 
                -lr LEARNING_RATE 
                -ts TEST_SIZE
                -dp DATASET_PATH
                -mp OUTPUT_MODEL_PATH
                -hd TRAINING_HISTORY_PATH

OR

python train.py --epochs NUMBER_OF_EPOCHS
                --batch-size BATCH_SIZE 
                --learning-rate LEARNING_RATE 
                --test-split TEST_SIZE
                --data-path DATASET_PATH
                --model-path OUTPUT_MODEL_PATH
                --history-dir TRAINING_HISTORY_PATH

recommended: USE THE DEFAULT TRAINING PARAMETERS USING:

python train.py

Test model

python test-model.py --help to show description of the parameters

test your model using the command below since the data gathering process takes alot of time, you can test our model thats in ./model

python test-model.py -mp MODEL_PATH

OR

python test-model.py --model-path MODEL_PATH

recommended: USE THE DEFAULT MODEL USING:

python test-model.py